Field of the Invention
The present invention relates to an image processing method for a picture photographed in an event, such as a marathon race.
Description of the Related Art
Conventionally, there has been known a technique for estimating a position of a race bib based on a detected position of a runner's face, and reading a race bib number using an OCR (Optical Character Reader) (see “Racing Bib Number Recognition” written by Idan Ben-Ami, Tali Basha, and Shai Avidan, http://www.eng.tau.ac.il/˜avidan/papers/RBNR.pdf).
However, the technique described in Non-PTL 1 has a problem that when a race bib number of a person is read from a photographed image, characters on e.g. a billboard or a road sign in the background within the image are erroneously detected as the race bib number. Further, in a case where a face of a person cannot be detected from an image, and in a case where a race bib is largely deformed, causing deformation of the shape of the race bib number, it is impossible to correctly read the race bib number by character recognition performed using the OCR. Further, a race bib attached to the body of a runner has a characteristic that in a case where the image is photographed from a lateral direction, the race bib is more largely deformed toward an end in a view depth direction, and a character interval is changed. The technique described in Non-PTL 1 assumes a case where a runner is photographed from the front, and hence it is impossible to correctly read the race bib number using the OCR in the above-described case.
Further, there is a problem that if a person overlaps the runner or if a hand of the runner is positioned in front of the race bib, part of the race bib is hidden, and this prevents the race bib number from being correctly recognized. If only part of the race bib number is detected in such a case, it is also difficult to determine whether or not the race bib number is correctly detected.